TopoPPI: A geometry-aware platform for visualization and low-distortion mapping of protein-protein interaction interfaces
Protein-protein interfaces encode molecular recognition through spatially organized contacts, yet their buried, curved surfaces are difficult to inspect as a whole. Flattening these surfaces can expose interface organization but introduces geometric distortion and may divide individual residues into disconnected regions. We present TopoPPI, a geometry-aware platform that converts interaction surfaces into low-distortion two-dimensional atlases while retaining links to source atoms, residues and contacts. TopoPPI extends OptCuts with a contact-weighted penalty for seam-induced fragmentation of residue footprints-the surface regions assigned to individual residues-jointly balancing continuity, seam length and distortion. In a leakage-controlled benchmark of 2,236 experimental complexes, the geometry-only OptCuts foundation achieved 88.39% lower family-weighted scale-fair distortion than the next-best tested parameterization across 2,209 complete comparable outputs. Across 2,204 exact-domain comparisons, TopoPPI reduced family-weighted residue-footprint fragmentation by 12.81% (95% confidence interval, 11.50-14.05%), with a 2.60% increase in normalized seam length and a 0.061% increase in scale-fair distortion. Continuity gains extended to 1,180 fixed-pose pairs of AlphaFold monomer models and 145 predicted complexes, with fragmentation reductions of 18.60% and 11.65%, respectively. Functionally annotated antibody and colicin-immunity interfaces demonstrated reconnection of the footprints of experimentally characterized binding and specificity residues. Further applications mapped published binding and folding free-energy estimates onto the KRAS-RAF1 interface and compared shared receptor residues across two growth-hormone recognition sites. Available through graphical and command-line interfaces, TopoPPI provides a residue-aware visualization framework for integrating structural contacts and functional annotations within geometrically controlled interface maps.
Authors
- Shuyu Zhong (ORCID: https://orcid.org/0009-0000-6218-1271)
Institutions
- Vrije Universiteit Amsterdam (NL)
- University of Amsterdam (NL)
Publication Details
- Journal
- PLoS Computational Biology
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1371/journal.pcbi.1014963
- Primary Topic
- Protein Structure and Dynamics
- Type
- article
- Field-Weighted Citation Impact
- 0.00